Metabolization of Flavan‐3‐ol Oligomers by Eggerthella lenta, a Human Gut Bacterium

ABSTRACT

Flavan‐3‐ols are among the most widely consumed flavonoids and have been implicated in the prevention of cardiometabolic diseases. Increasing evidence indicates that phenolic metabolites produced by the gut microbiota from flavan‐3‐ols contribute significantly to these effects. While some bacterial species can metabolize flavan‐3‐ol monomers, no isolates with activity on oligomers, i.e., procyanidins, have been reported. Since the gut microbiota is known to hydrolyze procyanidins, our aim was to identify bacteria capable of degrading them and to characterize their resulting metabolites. Through culturomic, genomic, and metabolomic approaches, we isolated novel strains of Eggerthella lenta and Flavonifractor plautii that metabolize flavan‐3‐ol monomers. Their activity was subsequently tested on B‐type (DP 2 to 4) and A‐type (DP 2) procyanidins. Only E. lenta was able to convert procyanidins, whereas F. plautii, either alone or in co‐culture with E. lenta, showed no detectable activity. The reaction catalyzed by E. lenta corresponded to C‐ring opening of the terminal unit of the oligomers. This work provides the first demonstration of procyanidin metabolism by a human gut bacterium. The presence and activity of E. lenta in the gut may therefore be a key determinant of the health benefits associated with flavan‐3‐ol–rich diets.

Novel human gut bacterial strains were isolated and characterized for their ability to metabolize flavan‐3‐ols. While Eggerthella lenta and Flavonifractor plautii metabolized flavan‐3‐ol monomers, only E. lenta converted B‐ and A‐type procyanidins, demonstrating that this gut bacterium can metabolize flavan‐3‐ol oligomers.

1. Introduction

Flavan‐3‐ols, a subclass of flavonoid polyphenols, are present in a wide range of plant‐based foods and beverages and are strongly recommended for the prevention of cardiometabolic diseases [1, 2, 3, 4]. Their global consumption ranges from 256 mg to 1800 mg per person per day [5]. In foods, they occur as aglycone monomers, oligomers, and polymers known as proanthocyanidins, or condensed tannins [6]. The most common monomers are (+)‐catechin and (−)‐epicatechin, followed by (−)‐epigallocatechin [6]. Procyanidins, a subgroup of proanthocyanidins composed of repeating (epi)catechin units, are the most widely consumed flavan‐3‐ols [5, 7]. Because of their oligomeric structure and their ability to form complexes with polysaccharides and proteins [8, 9], their bioavailability is very limited [10, 11]. Consequently, they reach the colon, where the commensal microbiota can convert them into metabolites that are potentially more bioavailable than their precursors [12, 13, 14, 15, 16].

The degradation pathway of flavan‐3‐ol monomers has been well described using in vitro fermentation with fecal microbiota combined with metabolomic approaches [17], although only a few of the bacteria involved have been isolated. The first steps of (epi)catechin degradation, consisting of C‐ring opening and B‐ring dehydroxylation, are catalyzed by several bacterial genera (Adlercreutzia, Gordonibacter, Eggerthella, Slackia) within the Eggerthellaceae family [18, 19, 20, 21, 22, 23, 24, 25]. These reactions generate metabolites that subsequently undergo A‐ring opening. To date, only one species, Flavonifractor plautii, has been shown to catalyze this step [19], leading to the formation of 5‐phenyl‐ɣ‐valerolactone (PVL) and 5‐phenylvaleric acid (PVA) derivatives, which are unique products of microbial flavan‐3‐ol catabolism [17]. To our knowledge, no microbial representatives have been identified as responsible for the subsequent degradation steps producing lower‐molecular‐weight phenolic acids such as phenylpropionic acid and phenylacetic acid derivatives. These metabolites are bioavailable [26, 27], and some of them—particularly PVL and PVA derivatives—are increasingly suspected of playing beneficial roles in health [28, 29, 30, 31, 32].

In contrast, the degradation pathway of procyanidins remains debated, further complicated by the existence of both A‐type and B‐type structures [12, 16, 28, 33, 34, 35, 36]. Studies using A2 and B2 dimers with complex fecal microbiota suggest that, as with monomers, C‐ring opening may be the first step of degradation [33]. Furthermore, in the presence of complex microbiota, procyanidins yield metabolites similar to those generated from monomers, indicating a shared degradation pathway [34, 35]. Thus, procyanidin oligomers may also contribute to beneficial health effects through the production of microbial metabolites. However, no intestinal bacteria have yet been identified as capable of degrading either A‐type or B‐type procyanidins. To address this knowledge gap regarding procyanidin microbial catabolism, the aim of this study was to identify bacteria able to metabolize low–degree of polymerization procyanidins using low‐throughput anaerobic culturomics, and to better elucidate the degradation pathway using metabolomic approaches.

2. Experimental Section

2.1. Molecules and Reagents

The (+)‐Catechin (≥98% purity) and (−)‐epicatechin (≥90% purity) were obtained from Sigma–Aldrich (Darmstadt, Germany). The A2‐type procyanidin dimer (≥99% purity) was purchased from Extrasynthèse (Lyon, France). B‐type procyanidins (dimer, trimer, and tetramer) were isolated from apple in this study. HPLC‐grade acetonitrile and methanol were obtained from VWR (Fontenay‐sous‐Bois, France), and HPLC‐grade acetic acid was obtained from Fisher Scientific (Pittsburgh, USA). Toluene‐α‐thiol was purchased from Sigma–Aldrich (Saint‐Quentin‐Fallavier, France).

2.2. Preparation and Quantification of Purified B‐Type Procyanidins (DP2–DP4) From Apple

The preparation and quantification of purified B‐type procyanidins (DP2–DP4) were performed as previously described [37].

Briefly, freeze‐dried apple powder from the Marie Ménard cider cultivar was sequentially extracted with hexane and methanol. The hexane extract was discarded. Methanol extracts containing procyanidins were pooled and concentrated using a rotary evaporator, followed by freeze‐drying. The dried methanol extract was purified first on a reverse‐phase silica gel Lichrospher 100 RP‐18 column (250 × 25 mm i.d., 5 µm; Merck, Darmstadt, Germany) to remove sugars, and then on a Luna HILIC 200 Å column (250 × 21.2 mm i.d., 5 µm; Phenomenex, Torrance, CA, USA) to isolate individual procyanidins. The dimer (PCB‐2) was collected between 13–17 min, the trimer (PCB‐3) between 20–24 min, and the tetramer (PCB‐4) between 25–29 min after gradient initiation. Each fraction (PCB‐2, PCB‐3, PCB‐4) was concentrated by rotary evaporation, freeze‐dried, and stored under vacuum at −20°C until use.

Procyanidins were analyzed by HPLC with diode array detection (DAD), with or without thioacidolysis, as described previously [38]. Analyses were performed using a Shimadzu Prominence ultra‐fast liquid chromatography system (Kyoto, Japan) controlled by LC Solution software (Shimadzu, Kyoto, Japan). Separations were carried out on a LiChroCART 250–4 Lichrospher 100 RP‐18 column (5 µm; Merck, Darmstadt, Germany) equipped with a LiChroCART 4‐4 Lichrospher RP‐18 guard cartridge (5 µm; Merck, Darmstadt, Germany), as described in [39].

The average degree of polymerization of procyanidins (𝐷𝑃̅𝑛) was calculated as the molar ratio of all flavan‐3‐ol units (thioether adducts plus terminal units, corrected for the naturally occurring (+)‐catechin and (−)‐epicatechin determined in samples without thiolysis) to the sum of (+)‐catechin and (−)‐epicatechin corresponding to terminal units (also corrected for their natural presence in samples without thiolysis).

2.3. Fecal Donors, Culture Conditions and Isolation of Flavan‐3‐ol Degrading Bacteria

Three fecal donors, hereafter referred to as O (49‐year‐old woman), P (26‐year‐old man), and T (27‐year‐old man), were healthy, omnivorous individuals. None had received antibiotics or other medications within three months prior to stool collection, knowingly consumed pre‐ or probiotics, or had a history of gastrointestinal or extra‐intestinal diseases. All donors were informed about the study objectives and procedures and provided written consent for the use of their fecal material in the experiments. This was a non‐interventional study with no deviation from routine clinical care. In accordance with French Public Health Law (CSP Art L 1121‐1.1), such protocols do not require ethics committee approval.

All microbial manipulations were performed in an anaerobic chamber (Jacomex, Lyon, France). Bacterial consortia and isolates were cultivated at 37°C in MCR20 medium, a complex medium derived from M10 medium [40]. The composition of MCR20 per liter was: 200 mL clarified rumen fluid, 75 mL mineral solution I, 75 mL mineral solution II, 10 mL volatile fatty acid solution, 0.5 g yeast extract, 2.0 g tryptone, 2 g glucose, 2 g cellobiose, 2 g maltose, 5 g arginine, 1 mL resazurin (0.1%), 5 mL hemin (0.1%), 2 mg vitamin K1, 0.5 g L‐cysteine hydrochloride, and 4 g NaHCO₃. Mineral solution I consisted of 6.0 g/L K2HPO4. Mineral solution II contained, per liter: 6.0 g KH₂PO₄, 12 g NH₄)₂SO₄, 12 g NaCl, 1.2 g MgSO₄, and 1.2 g CaCl₂. Rumen fluid was obtained from animals slaughtered at the INRAE experimental slaughterhouse (Saint–Genès–Champanelle, France) in accordance with local ethics committee guidelines and INRAE animal welfare regulations (Permit number: 63345001). The medium was adjusted to pH 6.9–7.0 and dispensed into Hungate tubes under CO₂. For agar media, 80 mL MCR20 medium was supplemented with 1.6 g agar (2% final concentration) in 100 mL anaerobic flasks. Tubes and flasks were autoclaved (120°C, 20 min) and stored at room temperature.

Flavan‐3‐ols were dissolved in methanol to prepare 50 mM stock solutions, aliquoted, and stored under CO₂ in dark vials at −20°C. They were added to MCR20 liquid or solid media immediately before use to achieve a final concentration of 0.5 mM.

Isolation of flavan‐3‐ol‐degrading bacteria from human feces was performed as described previously [19]. Briefly, successive enrichments were carried out with bacteria capable of degrading (+)‐catechin. Hundreds of isolates were then screened based on the disappearance or reduction of (+)‐catechin after 48 h of growth, monitored by HPLC–DAD using an Agilent 1200 HPLC system (Agilent, Les Ulis, France). The resulting isolates were named according to their source (O1: Donor O, stool 1; P1: Donor P, stool 1; P3: Donor P, stool 3; T1: Donor T, stool 1) followed by an arbitrary number and/or letter. They were further tested for catabolic activity against (−)‐epicatechin, the B‐type procyanidin oligomers prepared above, and a commercial A2‐type procyanidin dimer.

Bacterial cultures (500 µL) were diluted (2:3) in methanol/acetic acid (98/2, v/v) and centrifuged at 10 000 g for 10 min at 4°C. Supernatants were transferred to HPLC vials and stored at −20°C until analysis. For HPLC–DAD analysis, 5 µL of sample was injected onto a LiChroCART 4‐4 LiChrospher 100 RP‐8 pre‐column (4 × 4 mm i.d., 5 µm; Merck, Darmstadt, Germany) coupled to a LiChroCART 125‐4 Superspher 60 RP‐8 column (125 × 4 mm i.d., 4 µm; Merck, Darmstadt, Germany). The system was maintained at 35°C. Flavan‐3‐ol precursors and metabolites were separated with a mobile phase consisting of solvent A (water/formic acid, 100/0.05, v/v), solvent B (water/acetonitrile/formic acid, 30/70/0.05, v/v/v), and solvent C (pure acetonitrile), at a flow rate of 0.3 mL/min. The elution gradients used for monomers and procyanidins are detailed in Tables S1 and S2, respectively. A diode array detector (DAD) monitored absorbance at 275 nm for phenolic compounds. Data acquisition and processing were performed using Agilent OpenLab ChemStation software.

2.4. Structural Identification of Metabolites by UPLC–DAD–ESI/MS/MS Analyses

Flavan‐3‐ol metabolites were analyzed by LC–ESI–MS/MS using bacterial culture supernatants prepared as described above.

Separation and identification of procyanidin metabolites were performed on a Waters ACQUITY UPLC system (Milford, MA, USA) equipped with a UV–vis diode array detector and coupled to an HCT Ultra ion trap mass spectrometer (Bruker Daltonics, Bremen, Germany) with an electrospray ionization (ESI) source. Separations were achieved on an Acquity BEH C18 column (100 × 2.4 mm i.d., 1.7 µm; Waters). The column was maintained at 35°C, and 5 µL of sample was injected. The mobile phase was constituted by a binary solvent system with water/formic acid (99.95/0.05, v/v, solvent A,) and acetonitrile (solvent B); the elution gradient was as follows: 0–2 min, 5% B; 2–20 min, linear 5%‐70% B; 20–22 min, linear 70%‐90% B; 22‐22.5 min, 90%‐5% B; 22.5‐23.5 min, 5% B. The flow rate was set at 0.4 mL/min. Mass detection was conducted in the negative ionization mode from m/z 100 to 1200. MS conditions were as follows: capillary voltage of 3 kV, dry gas flow rate at 9 L/min, desolvation temperature at 365°C, and nebulization pressure at 50 psi. The mass spectra were then processed using Bruker Compass DataAnalysis v4.2 software (Bruker Daltonik, Bremen, Germany).

2.5. Extraction of Genomic DNA

Overnight pre‐cultures of Eggerthella lenta strains grown in MCR20 medium were used to inoculate (2% inoculum) 60 mL of the same medium. After 24 h of incubation, bacterial pellets were collected by centrifugation (10 000 g, 10 min, RT), washed once with 1 mL phosphate‐buffered saline (1 × PBS), and resuspended in 200 µL of 1× PBS after a second centrifugation. Cells were then mixed with 100 mg of 0.1 mm zirconium beads (BioSpec) and subjected to bead‐beating for 3 min at 30 Hz using a BeadBeater MM 400 (Retsch). Subsequent steps were performed using the Easy‐DNA kit (Invitrogen) according to the manufacturer's instructions, with modifications to obtain highly pure genomic DNA for sequencing. DNA was precipitated from the aqueous phase by adding one‐tenth volume of 3 M sodium acetate (pH 5.2) and two volumes of ice‐cold absolute ethanol (−20°C). The resulting DNA aggregate, suspended in ethanol, was recovered with a sterile Pasteur pipette, washed with 70% ethanol, air‐dried at room temperature for 10 min, and finally dissolved in 100 µL of 5 mM Tris/HCl buffer (pH 8.5). For PacBio sequencing, DNA purity was verified by comparing concentrations measured at 260 nm absorbance with those obtained using the Qubit dsDNA BR Assay Kit (Thermo–Fisher Scientific, Lyon, France). A ratio of 2 was required.

2.6. Genome Library Preparation and Sequencing

Single‐molecule real‐time (SMRT) sequencing was performed at the Gentyane Sequencing Platform (Clermont‐Ferrand, France) using a PacBio Sequel II sequencer (Pacific Biosciences, Menlo Park, CA, USA). SMRTbell libraries were prepared with the SMRTbell Express Template Prep Kit 2.0, following the Procedure and Checklist: Preparing Multiplexed Microbial Libraries Using SMRTbell Express Template Prep Kit 2.0. Genomic DNA (1 µg per strain) was sheared using g‐tubes (Covaris, England) to generate fragments of approximately 10 kb. Fragment size distribution was assessed with a Fragment Analyzer (Agilent Technologies, Santa Clara, CA, USA). Sheared DNA was then subjected to enzymatic treatments to remove single‐strand overhangs and repair potential backbone damage. An A‐tailing reaction followed by ligation of barcoded overhang adapters was performed to produce SMRTbell templates. Barcoded samples were pooled in equimolar amounts and purified with 0.45 × AMPure PB beads to obtain final libraries of ∼10 kb. Library quality and concentration were assessed using a Fragment Analyzer (Agilent Technologies) and a Qubit fluorometer with the Qubit dsDNA HS Assay Kit (Life Technologies). A sequencing‐ready SMRTbell–polymerase complex was prepared with Binding Kit 2.0 (PacBio) and primer V4, using the diffusion loading protocol according to the manufacturer's instructions. The PacBio Sequel II instrument was programmed to load a 100 pM library and sequenced in CLR mode on a SMRT Cell 8M (PacBio) using Sequencing Plate 2.0, with a 2 h pre‐extension and acquisition of one 15 h movie per SMRT Cell.

2.7. Assembly and Analysis of Bacterial Genomes

Genome assembly was performed by the Gentyane platform (INRAE, Clermont–Ferrand, France). Raw sequences were demultiplexed using lima (v1.11.0). De novo genome assembly was carried out with the “Microbial Assembly” tool of the SMRT Link software suite (v8.0, PacBio), following the manufacturer's recommendations. Genome completeness was assessed with CheckM (v1.1.3) using default parameters [41], and genome annotation was performed with Prokka (v1.14.5) [42]. Average nucleotide identity (ANI) was calculated using OrthoANI (v0.93.1) [43], which measures the nucleotide‐level similarity across orthologous genomic regions. ANI is widely used as an index of genomic relatedness, with values >95% generally indicating that two strains belong to the same species. The closer the ANI is to 100%, the more genetically similar the strains are. In this study, isolates from the same sample with ANI values >99.99% were considered to represent the same strain. Pangenome analysis was conducted with Roary (v3.11.2) [44] to assess genetic diversity and relatedness among sequenced genomes, identifying both the core genome (genes shared across isolates) and the accessory genome (isolate‐specific genes). The genomes used for this analysis were retrieved from the NCBI database (https://www.ncbi.nlm.nih.gov/). They are listed with their accession numbers in Tables S5 and S6. The presence of plasmids was predicted using mlplasmids (v1.0.0) [45] and searched in NCBI database using BLASTn against whole‐genome shotgun contigs (wgs) of Flavonifractor. All sequence data of this study are available in the NCBI Genome databases under BioProject accession number PRJNA1418111.

3. Results

3.1. Isolation of Human Gut Bacterial Strains Degrading Flavan‐3‐ol Monomers

The overall strategy was to isolate bacteria capable of metabolizing flavan‐3‐ol monomers and subsequently evaluate their ability to metabolize flavan‐3‐ol oligomers. To this end, four bacterial isolation campaigns were conducted using stool samples from three individuals (O, P, and T), including two campaigns from the same donor performed 2 months apart (designated P1 and P3).

Following successive culture enrichments, 380 bacterial isolates were screened for their ability to degrade (+)‐catechin by monitoring its disappearance from the culture medium after 24 h of incubation. Twenty isolates from the three donors were found to actively metabolize (+)‐catechin. Sequencing of the 16S rRNA gene of these isolates revealed that they all belonged to the species E. lenta (data not shown).

3.2. Highlighting of Two Metabolic Groups Within E. lenta Isolates

HPLC–DAD chromatograms revealed that isolates from stools P3 and T produced a metabolite (Figure 1A,B), identified by LC–ESI–MS/MS as 1‐(3′,4′‐dihydroxyphenyl)‐3‐(2″,4″,6″‐trihydroxyphenyl)propan‐2‐ol (3′,4′‐diOHPP‐2‐ol) from (+)‐catechin (Table 1). In contrast, isolates from stools O and P1 generated not only this metabolite but also an additional compound (Figure 1C), identified as 1‐(3′‐ or 4′‐hydroxyphenyl)‐3‐(2″,4″,6″‐trihydroxyphenyl)propan‐2‐ol (3′‐ or 4′‐OHPP‐2‐ol) (Table 1).

FIGURE 1.

Metabolization of (+)‐ catechin and (−)‐ epicatechin by E. lenta isolates. (A) HPLC–DAD chromatograms at 275 nm after 48 h incubation at 37°C of sterile MCR20 culture medium (blue) with (+)‐catechin (red) or (–)‐epicatechin (green). (B) HPLC–DAD chromatograms at 275 nm after 48 h incubation at 37°C of MCR20 culture medium with (+)‐catechin (red), (–)‐epicatechin (green), (+)‐catechin + E. lenta P3‐J7.5 (blue), and (–)‐epicatechin + E. lenta P3‐J7.5 (pink). (C) HPLC–DAD chromatograms obtained as in panel B, except that the bacterial isolate tested was E. lenta O1.56. (+)‐Catechin (C) and its two metabolites, 3’,4’‐diOHPP‐2‐ol and 3’‐OHPP‐2‐ol, eluted at 16.4, 17.3, and 19.7 min, respectively. (–)‐Epicatechin (EC) and its two metabolites, 3’,4’‐diOHPP‐2‐ol and 3’‐OHPP‐2‐ol, eluted at 16.8, 17.3, and 19.7 min, respectively. E. lenta isolates P3 and T produced chromatograms identical to those in panel B. E. lenta isolates O and P1 produced chromatograms identical to those in panel C. The E. lenta isolates and their genomic features are listed in Table S3.

TABLE 1.

Identification of (+)‐catechin and its metabolites by LC–ESI–MS/MS in cultures of E. lenta alone or in coculture with F. plautii.

Molecule	[M–H]− (m/z)	MS2 fragments (m/z)	Identified compounds	Abbreviations
(+)‐Catechin	289	271, 245, 231
E. lenta metabolites
	291	247, 205, 167	1‐(3’,4’‐dihydroxyphenyl)‐3‐(2'',4'',6''‐trihydroxyphenyl) propan‐2‐ol	3’,4’‐diOHPP‐2‐ol
	275	231, 217, 191	1‐(3’or 4’‐hydroxyphenyl)‐3‐(2'',4'',6''‐trihydroxyphényl) propan‐2‐ol	3’or 4’‐OHPP‐2‐ol
E. lenta + F. plautii metabolites
	225	207, 163, 123, 101	4‐hydroxy‐5‐(3',4'‐dihydroxyphenyl)‐valeric acid	3’,4’‐diOHPVA
	207	163, 122	5‐(3',4'‐dihydroxyphenyl)‐y‐valerolactone	3’,4’‐diOHPVL
	209	191, 147, 107, 101	4‐hydroxy‐5‐(3' or 4’‐hydroxyphenyl)‐valeric acid	3’or 4’‐OHPVA
	191	147	5‐(3'‐hydroxyphenyl)‐y‐valerolactone	3’or 4’‐OHPVL

Note: All the metabolites were identified by comparison with published data [34, 46]. 3',4'‐diOHPP‐2‐ol was produced by all E. lenta isolates while 3’or 4’‐OHPP‐2‐ol were only produced by E. lenta isolates O and P1. 3',4'‐diOHPVA and 3',4'‐diOHPVL were produced with (E. lenta P3‐J7.5/F. plautii P3‐I6) coculture from 3',4'‐diOHPP‐2‐ol. 3' or 4'‐OHPVA and 3' or 4’‐OHPVL were produced with (E. lenta O1‐56 / F. plautii P3‐I6) coculture from 3' or 4'‐OHPP‐2‐ol. Major fragment ions are in bold.

The MS/MS fragmentation patterns of these two metabolites matched previously reported data [34, 46]. Although mass spectrometry cannot distinguish whether dehydroxylation occurs at the 3′ or 4′ position of the B ring, several E. lenta strains have been described as catalyzing only 4′‐dehydroxylation of several flavan‐3‐ol monomers [18, 20, 22, 46, 47, 48]. On this basis, we infer that the dehydroxylated metabolite observed here corresponds to 1‐(3′‐hydroxyphenyl)‐3‐(2″,4″,6″‐trihydroxyphenyl)propan‐2‐ol (3′‐OHPP‐2‐ol) (Figure 1C).

Comparable results were obtained with (–)‐epicatechin, the enantiomer of (+)‐catechin. Isolates from stools P3 and T produced a single metabolite, whereas isolates from O and P1 generated two metabolites (Figure 1). These compounds exhibited identical retention times and UV spectra to those derived from (+)‐catechin. Consistent with earlier reports [22, 34], LC–MS/MS confirmed them as the same molecules, although the chirality of the metabolites cannot be determined by this method.

In summary, all 20 E. lenta isolates were capable of C‐ring opening of both (+)‐catechin and (–)‐epicatechin. However, they segregated into two distinct metabolic groups depending on their ability to further dehydroxylate the B ring: only the O1 and P1 isolates performed this second reaction.

3.3. Genomics of E. lenta Isolates

Our culturomic strategy yielded 20 bacterial isolates of the species E. lenta from three individuals (O, P, T). Since multiple isolates originating from the same stool sample could correspond to identical strains, it was essential to precisely affiliate them in order to reduce redundancy in downstream analyses. For this purpose, we performed whole‐genome sequencing of all 20 isolates using PacBio Sequel II technology. This third‐generation platform is particularly well suited for high‐quality bacterial genome assemblies due to its ability to generate long reads with low error rates, enabling reliable strain‐level resolution.

Except for one isolate (T1E11), sequencing enabled reconstruction of the genomes into a single contig (circular chromosome) for 15 isolates and into two contigs for 4 isolates (Table S3). Genome completeness, assessed using CheckM (v1.1.3), was excellent (99%–100%). Despite shared genomic features across isolates (GC content, rRNA and tRNA gene copy numbers), substantial variation was observed in genome size (3.28–3.65 Mb) and coding sequence counts (2780–3136) (Table S3).

Comparative genomic analyses based on Average Nucleotide Identity (ANI) and pangenome profiling revealed that isolates from the same stool sample (O, P1, P3, T) generally shared a very close genetic background, confirming they belonged to the same strain. One exception was isolate P1‐35, which was genetically closer to O1 than to other P1 isolates (data not shown). These findings were further supported by comparisons with 60 publicly available E. lenta genomes (Figure S1), which also demonstrated that our P1 isolates are closely related to the type strain DSM2243.

Based on these genomic insights, and in order to further explore the metabolic activities of our isolates toward flavan‐3‐ol oligomers, we selected two phylogenetically distinct representatives of the two metabolic groups described above: O1‐56 and P3‐J7.5.

3.4. Isolation of Human Gut Bacterial Strains Degrading E. lenta Metabolites

To identify microorganisms capable of further metabolizing the catabolites produced by E. lenta, we designed an isolation strategy based on culture supernatants enriched in 1‐(3',4'‐dihydroxyphenyl)‐3‐(2'',4'',6''‐trihydroxyphenyl)propan‐2‐ol (3',4'‐diOHPP‐2‐ol), since this metabolite is not commercially available. A single isolation campaign was performed using stool sample P3. From 150 isolates obtained after enrichment, only two showed activity toward 3',4'‐diOHPP‐2‐ol. Sequencing of the 16S rDNA gene identified both isolates as belonging to the species F. plautii (data not shown).

3.5. Genomics of F. plautii Isolates

Complete genome sequencing of our two F. plautii isolates revealed a high degree of genetic similarity (Table S4). Both isolates harbored a 3.9 Mb genome and an additional contig of ∼46 kb, whose sequence was 100% identical between them. The mlplasmids software (v1.0.0) predicted this contig to be a plasmid with a probability of 0.997. A BLASTn search against wgs NCBI database showed that this plasmid is present in other strains of F. plautii.

Pangenome analysis showed that the two isolates shared 3758 CDS (99.9%), supporting the hypothesis that they represent the same strain. However, the ANI value between the genomes was 99.975%, slightly below our predefined threshold of 99.99%. This result was further confirmed by comparison with 42 publicly available F. plautii genomes (Figure S2), which also highlighted that our isolates are phylogenetically close to the F. plautii type strain DSM4000.

Based on these genomic findings, we selected isolate F. plautii P3‐I6 as the representative strain for subsequent metabolic analyses on flavan‐3‐ols.

3.6. Metabolization of Flavan‐3‐ol Monomers by (E. lenta and F. plautii) Co‐Cultures

Analysis of HPLC‐DAD chromatograms showed that F. plautii P3‐I6 was unable to degrade (+)‐catechin or its enantiomer (−)‐epicatechin (data not shown). However, when co‐cultured with E. lenta P3‐J7.5, which converts these monomers into 3',4'‐diOHPP‐2‐ol (Figure 1B), two additional metabolites were detected. LC–ESI–MS/MS identified them as 4‐hydroxy‐5‐(3',4'‐dihydroxyphenyl)‐valeric acid (3',4'‐diOHPVA) and its lactone derivative, 5‐(3',4'‐dihydroxyphenyl)‐γ‐valerolactone (3',4'‐diOHPVL). Only 3',4'‐diOHPVA was visible on chromatograms, as 3',4'‐diOHPVL co‐eluted with a medium component (Table 1; Figure 2A,B).

FIGURE 2.

Metabolization of (+)‐catechin and (−)‐epicatechin by cocultures of E. lenta and F. plautii. (A) HPLC‐DAD chromatograms at 275 nm after 48 h of culture of (+)‐catechin alone (red), (+)‐catechin + E. lenta P3‐J7.5 (blue), and (+)‐catechin + E. lenta P3‐J7.5 + F. plautii P3.I6 (green). (B) HPLC–DAD chromatograms obtained as in panel A, except that (+)‐catechin was replaced by (–)‐epicatechin. (C) HPLC–DAD chromatograms obtained as in panel A, except that the bacterial isolate tested was E. lenta O1.56. (D) HPLC–DAD chromatograms obtained as in panel B, except that the bacterial isolate tested was E. lenta O1.56. (+)‐Catechin (C) and its metabolites 3’,4’‐diOHPP‐2‐ol and 3’,4’‐diOHPVA eluted at 16.4, 17.3, and 15.7 in, respectively. (–)‐Epicatechin (EC) and its metabolites 3’,4’‐diOHPP‐2‐ol and 3’,4’‐diOHPVA eluted at 16.8, 17.3, and 15.7 min, respectively. In all panels, the 3’,4’‐diOHPVL peak was not visible because it co‐eluted with a medium component at 19.3 min (retention time confirmed using 3’,4’‐diOHPVL as a standard and by UPLC–DAD‐ESI/MS/MS; see Table 1). Similarly, in panels C and D, the 3’‐OHPVL peak was not visible, although it was detected by UPLC‐DAD–ESI/MS/MS (see Table 1).

Furthermore, F. plautii P3‐I6 was able to transform 3'‐OHPP‐2‐ol, produced by E. lenta O1‐56 from (+)‐catechin or (‐)‐epicatechin (Figure 1C), into 3'‐OHPVA and 3'‐OHPVL. Although not detected in chromatograms (Figure 2C,D), these two metabolites were present in trace amounts in the culture medium since they were identified by mass spectrometry (Table 1).

In summary, these results, consistent with previous reports [34, 46], demonstrate that F. plautii P3‐I6 opens the A‐ring of both 3',4'‐diOHPP‐2‐ol and 3'‐OHPP‐2‐ol, initially generated by E. lenta, leading to the formation of hydroxyphenylvaleric acid and hydroxyphenyl‐γ‐valerolactone derivatives.

3.7. Characterization of Purified B‐Type Procyanidins From Apple

Three procyanidin fractions, designated PCB‐2, PCB‐3, and PCB‐4 (see Materials and Methods; Figure S3), were purified and further characterized.

The PCB‐2 fraction contained more than 85% phenolic compounds and was composed primarily of the B2 dimer [(‐)‐epicatechin‐4→8‐(−)‐epicatechin], with the B1 dimer [(−)‐epicatechin‐4→8‐(+)‐catechin] accounting for ∼7% of total procyanidins. Thioacidolysis revealed that (−)‐epicatechin represented 95% of the constitutive units, while (+)‐catechin, present exclusively as a terminal unit, represented only 3.5% (Table 2). These results were consistent with direct HPLC–DAD quantification and with previous findings by Watrelot et al. [37]. Minor compounds, including phloretin xyloglucoside and 5‐caffeoylquinic acid, accounted for  < 1% of the fraction.

TABLE 2.

Composition of the B‐type procyanidin fractions purified from apple.

	Total procyanidins	Procyanidin characteristics				PLX	5‐CQA	Total polyphenols
			(+)‐Catechin (terminal unit)	(‐)‐Epicatechin (terminal unit)	(‐)‐Epicatechin (extension unit)
Fractions	(mg/g DM)	Calculated mDP	% Total procyanidins			(mg/g DM)
PCB‐2	847	2.0	5	46	49	2.2	1.1	850.3
PCB‐3	750	2.7	3	34	63	0.2	0.4	750.6
PCB‐4	753	3.4	3	27	70	0	0.2	753.2

Abbreviations: PLX, phloridzin xyloglucoside; 5‐CQA, 5‐O‐Caffeoylquinic acid; DM, dry matter; mDP, mean degree of polymerization.

The PCB‐3 fraction contained more than 75% phenolic compounds and consisted mainly of DP3 procyanidin isomers. Thioacidolysis showed that (−)‐epicatechin accounted for 97% of the constitutive units, while (+)‐catechin (only as a terminal unit) accounted for ∼3% (Table 2). Trace amounts ( < 0.5%) of phloretin xyloglucoside and 5‐caffeoylquinic acid were also detected.

The PCB‐4 fraction contained more than 75% phenolic compounds and consisted of DP4 procyanidins. (−)‐Epicatechin was the predominant unit, present both as extension units (70%) and as terminal units (27%), while (+)‐catechin occurred only as a terminal unit, representing ∼3% of the constitutive units (Table 2). Trace contaminants, including phloretin xyloglucoside and 5‐caffeoylquinic acid, represented  < 0.1% of the fraction.

3.8. Evidence of Procyanidin Metabolization by E. lenta

The catalytic activities of E. lenta and F. plautii isolates were assessed using B‐type procyanidin dimers (PCB‐2), trimers (PCB‐3), and tetramers (PCB‐4) prepared as described above. HPLC–DAD analyses showed that all four E. lenta strains (O1‐56, P1‐71, P3‐J7.5, T1‐E1) converted PCB‐2 into a metabolite termed PCB‐2M (Figure 3A). Similar conversions occurred with trimers and tetramers, yielding PCB‐3M and PCB‐4M, respectively (Figure 3B,C). By contrast, F. plautii P3‐I6 displayed no catalytic activity on any of the procyanidin oligomers or their corresponding E. lenta metabolites (Figure 3).

FIGURE 3.

Metabolization of B‐type procyanidins (DP2 to 4) by E. lenta. (A) HPLC‐DAD chromatograms at 275 nm after 48 h of culture with the B‐type procyanidin dimer fraction (PCB‐2), in the absence of bacteria (top, blue), in the presence of E. lenta P3‐J7.5 alone (top, red), F. plautii P3‐I6 alone (bottom, blue), or E. lenta P3‐J7.5 + F. plautii P3‐I6 (bottom, red). PCB‐2 and its metabolite (PCB‐2M) eluted at 11.0 and 12.7 min, respectively. (B) HPLC‐DAD chromatograms at 275 nm after 48 h of culture with the B‐type procyanidin trimer fraction (PCB‐3) under the same conditions as in panel A. PCB‐3 and its metabolite (PCB‐3M) eluted at 11.5 and 12.2 min, respectively. (C) HPLC–DAD chromatograms at 275 nm after 48 h of culture with the B‐type procyanidin tetramer fraction (PCB‐4) under the same conditions as in panel A. PCB‐4 and its metabolite (PCB‐4M) eluted at 11.7 and 12.4 min, respectively. Similar chromatograms (panels A–C) were obtained with the three other E. lenta strains (O1‐56, P1‐71, T1‐E1), either alone or in coculture with F. plautii P3‐I6 (not shown).

To further probe structural specificity, two E. lenta strains (O1‐56 and P3‐J7.5) were tested with a commercial A‐type procyanidin dimer (PCA‐2). Unlike B‐type dimers, PCA‐2 possesses an additional ether bond (C2–O–C7 or C2–O–C5) (Figure 6). Consistent with earlier observations [49], PCA‐2 underwent spontaneous isomerization under incubation conditions (48 h, 37°C, pH 6.9, anaerobiosis), producing two additional isomers of identical MW but distinct fragmentation profiles (Figure 4). Despite this, a single metabolite designed PCA‐2M, could arise from incubation with both E lenta strains.

FIGURE 6.

Schematic representation of the degradation of B‐type and A‐type procyanidin dimers by E. lenta. C‐ring opening (blue arrow) by E. lenta strains O1‐56, P1‐71, P3‐J7.5, T1‐E1 occurs on the terminal unit of the dimers. PCB‐2 = B‐type procyanidin dimer, PCB‐2M = metabolite of PCB‐2, PCA‐2 = A2‐type procyanidin dimer, PCA‐2M: metabolite of PCA‐2.

FIGURE 4.

Metabolization of A‐type procyanidins (DP2) by E. lenta. HPLC–DAD chromatograms at 275 nm after 48 h of culture with the A2‐type procyanidin dimer (PCA‐2), in the absence of bacteria (blue), or in the presence of E. lenta O1‐56 (red) or E. lenta P3‐J7.5 (green). Under our culture conditions, PCA‐2 formed three isomers in the absence of bacteria; the initial form is marked with an asterisk. The three isomers and the bacterial metabolite (PCA‐2M) eluted at 13.5, 14.0*, 14.3, and 15.2 min, respectively.

The four metabolites (PCB‐2M, PCB‐3M, PCB‐4M, PCA‐2M) were characterized by LC–ESI–MS/MS (Table 3).

TABLE 3.

LC–ESI–MS/MS fragmentation of the flavan‐3‐ol oligomers used in this study and their corresponding metabolites produced by E. lenta (strains P3‐J7.5 and O1‐56).

Molecules	[M–H]− (m/z)	MS2 fragments (m/z)	MS3 fragments (m/z)
PCB‐2	577	559, 451 (HRF fragment), 425 (RDA fragment), 407 (425 – H2O), 289, 287, 245	MS (407): 389, 339, 297, 285, 281, 257, 255, 243
PCB‐2M	579	561, 453 (HRF fragment), 427 (RDA fragment), 409, 291 (open lower EC unit)	MS (409): 391, 341, 285, 269, 243
PCB‐3	865	847, 739 (HRF fragment), 713 (RDA fragment), 695 (713 – H2O), 577, 575, 543, 451, 407, 363, 289, 287	MS (695): 677, 543, 525, 451, 407, 363, 289, 243
PCB‐3M	867	849, 741 (HRF fragment), 715 (RDA fragment), 697 (715 – H2O), 621, 579, 575, 453, 409, 291 (open lower EC unit), 287 (upper EC unit)	MS (697): 679 (‐H2O), 571, 527 (679‐ RDA fragment), 453, 409, 291 (open lower EC unit), 243
PCB‐4	1153 [M–H]− 576 [M‐2H]2−	MS (1153): 1135, 1027 (HRF fragment), 1001 (RDA fragment), 983, 907, 865, 863, 739, 575, 449, 407; MS (576): 1027 (HRF fragment), 863 (‐EC), 739, 701, 559, 531, 500, 491, 451, 449, 407, 289, 287, 245	—
PCB‐4M	1155 [M–H]− 577 [M‐2H]2−	MS (1155): NF; MS (577): 1029 (HRF fragment), 863 (‐EC+2H), 741, 701, 501, 492, 453, 413, 409, 361, 333, 291 (open lower EC unit), 287 (upper EC unit), 243	—
PCA‐2	575	557, 539,449 (HRF fragment), 423 (RDA fragment),327, 289, 287, 285, 245	MS (423): 405, 313, 297 (‐HRF frag.), 285 (oxidized upper EC moiety), 257, 241, 137
PCA‐2M	577	559, 541, 451 (HRF fragment), 409, 329, 291 (open lower EC unit), 285 (oxidized upper EC unit)	—

Note: PCB‐2, PCB‐3 and PCB4 correspond to the purified fractions of B‐type procyanidins enriched in dimers, trimers and tetramers, respectively. PCB‐2M, PCB‐3M, and PCB‐4M are their corresponding metabolites. PCA‐2 is a commercial A2‐type procyanidin dimer, and PCA‐2M is its corresponding metabolite. Major fragment ions are in bold, next intense ions are underlined.

Abbreviations: HRF, heterocyclic ring fission; RDA, Retro–Diels–Alder reaction; EC, (Epi)Catechin.

3.8.1. Structural Elucidation of PCB‐2M

The parent ion of PCB‐2M was detected at m/z 579, representing +2 Da compared to the precursor dimer (PCB‐2, m/z 577). This shift indicated reduction of one flavan‐3‐ol unit. MS² analysis revealed fragments at m/z 561 (−H₂O), 453 (HRF, Heterocyclic Ring Fission), 427 (RDA, Retro–Diels–Alder reaction, −152 Da), 409 (further −H₂O), and 291 (monomeric unit after interflavan bond cleavage). The most intense fragment (m/z 409) pointed to a predominant RDA pathway, consistent with a C‐ring open structure at the terminal unit. Supporting this, the diagnostic fragment at m/z 291 could only arise from interflavan bond cleavage of a dimer in which the lower (terminal) unit had undergone C‐ring opening. If the upper unit were modified instead, the expected fragment would have appeared at m/z 289. The MS³ profile (m/z 409 → 391, 341, 285, 269, 243) included the two most intense ions (285 and 243) reported by Stoupi et al. [33]. PCB‐2M was therefore identified as a C‐ring open form at the terminal unit.

3.8.2. Structural Elucidation of PCA‐2M

The PCA‐2M metabolite displayed a parent ion at m/z 577 (+2Da relative to PCA‐2, m/z 575). A key difference from PCA‐2 was the disappearance of the diagnostic fragment at m/z 423 (generated by RDA of the lower unit), consistent with C‐ring opening at the lower unit of PCA‐2M. Additional fragments strongly supported this interpretation:

1. HRF fragment shifted from m/z 449 (PCA‐2) to m/z 451 (PCA‐2M), consistent with HRF restricted to the upper unit for PCA‐2 and PCA‐2M.

2. Quinone methide fragmentation followed by interflavan bond cleavage yielded m/z 291 (major; terminal unit, C‐ring open) and m/z 285 (minor; upper unit).

3. Absence of m/z 287/289 ruled out reduction of the upper unit.

The fragmentation profile matched that of PCA‐2 incubated with pig fecal microbiota [13], confirming C‐ring opening at the terminal unit, as also observed for PCB‐2.

3.8.3. Structural Elucidation of PCB‐3M and PCB‐4M

For PCB‐3M, the parent ion appeared at m/z 867 (+2Da relative to precursor, m/z 865). Key fragments included m/z 741 (HRF) and m/z 715 (RDA), with subsequent dehydrations yielding m/z 697 and m/z 679. Interflavan bond cleavage provided diagnostic ions at m/z 579/287 (upper–central cleavage) and m/z 575/291 (central–lower cleavage). In contrast, alternative reduced structures would have produced m/z 577/289. Thus, PCB‐3M was identified as the C‐ring open form at the terminal unit.

Similarly, PCB‐4M presented a parent ion at m/z 1155 (+2Da relative to precursor, m/z 1153). Fragments at m/z 863/291 support well an interflavan bond cleavage between an intermediate unit and the lower one with a C‐ring open form at the terminal unit.

4. Discussion

Using a culturomic strategy based on successive enrichments and monitoring the catabolism of (+)‐catechin by HPLC–DAD, we isolated twenty E. lenta strains and two F. plautii strains from the fecal microbiota of three healthy individuals. Genome sequencing followed by comparative analysis allowed us to eliminate genetically close, likely clonal isolates, reducing the collection to four representative E. lenta strains (O1‐56, P1‐71, P3‐J7.5, T1‐E1) and one F. plautii strain (P3‐I6).

LC–ESI–MS/MS analyses of culture supernatants demonstrated that all four E. lenta strains cleaved the C‐ring of (+)‐catechin and (−)‐epicatechin, producing 3′,4′‐diOHPP‐2‐ol. Two strains (O1‐56, P1‐71) further catalyzed B‐ring dehydroxylation. While MS does not discriminate between loss of the 3′‐ or 4′‐hydroxyl group, previous studies consistently showed that 3′‐OHPP‐2‐ol is formed by other E. lenta strains [18, 20, 22, 46, 47, 48]. The F. plautii strain P3‐I6 was then able to open the A‐ring of these intermediates, generating 3′,4′‐diOHPVA and 3′,4′‐diOHPVL from 3′,4′‐diOHPP‐2‐ol, as well as 3′‐OHPVA and 3′‐OHPVL from 3′‐OHPP‐2‐ol (Figure 5). Although stereochemistry could not be determined by LC–ESI–MS/MS, it is likely that the metabolites retained the chirality of their precursors [22].

FIGURE 5.

Schematic representation of (+)‐catechin and (−)‐ epicatechin metabolization by the E. lenta and F. plautii strains isolated in this study. C‐ring opening (blue arrow) by E. lenta strains O1‐56, P1‐71, P3‐J7.5, and T1‐E1 produces 1‐(3′,4′‐dihydroxyphenyl)‐3‐(2′′,4′′,6′′‐trihydroxyphenyl)propan‐2‐ol (3′,4′‐diOHPP‐2‐ol). This metabolite undergoes dehydroxylation on the B‐ring by E. lenta strains O1‐56 and P1‐71 only. According to the literature [18, 20, 22, 46, 47, 48], this reaction occurs at the 4′ carbon (red arrow), yielding 1‐(3′‐hydroxyphenyl)‐3‐(2′′,4′′,6′′‐trihydroxyphenyl)propan‐2‐ol (3′‐OHPP‐2‐ol). A‐ring opening (green arrow) by F. plautii strain P3‐I6 generates additional metabolites from the E. lenta products: 3′,4′‐diOHPVA = (+)‐ 4‐hydroxy‐5‐(3′,4′‐dihydroxyphenyl) valeric acid; 3′,4′‐diOHPVL = (+)‐ 5‐(3′,4′‐dihydroxyphenyl)‐γ‐valerolactone; 3′‐OHPVA = (+)‐ 4‐hydroxy‐5‐(3′‐hydroxyphenyl) valeric acid; 3′‐OHPVL = (+) 5‐(3′‐hydroxyphenyl)‐γ‐valerolactone.

These findings are consistent with prior reports, as E. lenta and F. plautii are among the species involved in flavan‐3‐ol monomer metabolism [19, 22, 25]. Importantly, our study experimentally confirmed that B‐ring dehydroxylation is strain‐dependent within E. lenta, in agreement with genomic data indicating that only subsets of strains harbor the cadh operon [48].

Our approach illustrates both the specificity and the limitations of targeted culturomics. The isolation of strains already described in the literature suggests that the nutrient‐rich medium used may still lack essential cofactors required by less readily cultivable bacteria. In contrast, high‐throughput culturomics, as pioneered by Lagier et al. [50], employs diverse culture media and robotics to streamline colony picking and phylogenetic screening, enabling isolation of  > 1000 prokaryotic species, ∼20% of which are potentially novel. By comparison, our more focused approach, starting from three individuals, produced 380 isolates, which were screened by HPLC‐DAD for flavan‐3‐ol catabolism. A drawback of high‐throughput culturomics, however, is the lack of equally high‐throughput phenotype screening methods for specialized metabolisms such as polyphenol degradation. Although such strategies exist for dietary fiber metabolism [51], none have yet been adapted to polyphenols.

To study microbial metabolism of flavan‐3‐ol oligomers, sufficient substrate quantities were required. Because very few procyanidins are commercially available (limited to B2 and A2 dimers), we purified B‐type procyanidins from apple parenchyma (cv. Marie Menard). Three fractions (∼500 mg each) were obtained, enriched in dimers (mDP = 2.0), trimers (mDP = 2.7), and tetramers (mDP = 3.4), primarily composed of (−)‐epicatechin subunits.

When tested on these substrates, all four E. lenta strains catalyzed the conversion of procyanidins. LC–ESI–MS/MS identification revealed that B‐type dimers, trimers, tetramers, and the A2 dimer underwent C‐ring opening at the terminal unit, consistent with the pathway previously observed for (−)‐epicatechin (Figure 6). The metabolite PCA‐2M formed from dimer A2 matched that reported by Engemann et al. (2012) following incubation with pig fecal microbiota, while PCB‐2M corresponded to one of the 24 metabolites described in the human fecal incubation of dimer B2 [33]. Notably, unlike Stoupi et al., we did not detect metabolites with both C‐rings opened, suggesting that E. lenta performs the initial step of procyanidin degradation but does not advance further. Similarly, our F. plautii isolate displayed no activity on B‐type dimers or on the PCB‐2M metabolite, implying that its role may come later in the degradation cascade, potentially once both C‐rings have been cleaved.

This study establishes E. lenta as a key initiator of procyanidin metabolism in the human gut by catalyzing C‐ring opening at the terminal unit of both B‐type and A‐type procyanidins. Our results support the use of E. lenta as a bacterial marker for intestinal (epi)catechin and procyanidin catabolism. Further research is needed to identify additional microbial partners responsible for downstream transformations leading to smaller, bioavailable phenolic metabolites of potential benefit to the host.

Funding

This work was funded by the INRAE MICA department (France) and HerbalT private company (France).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: mnfr70615‐sup‐0001‐SuppMat.pptx.

Supporting File 2: mnfr70615‐sup‐0002‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are openly available in Prokaryotic and eukaryotic genomes at http://www.ncbi.nlm.nih.gov/bioproject/1418111, reference number PRJNA1418111.